Prevent ABI changes affect EnzymeAD
This PR handles ABI changes for autodiff input arguments to improve Enzyme compatibility. Fundamentally this adjusts activities when a function argument is lowered as an `ScalarPair`, so there's no mismatch between diff activities and args. Also removes activities corresponding to ZSTs.
fixes: https://github.com/rust-lang/rust/issues/144025
r? `@ZuseZ4`
Lint more overlapping assignments in MIR.
In an effort to make bugs like https://github.com/rust-lang/rust/issues/146383 more easily discovered, this PR extends the "overlapping assignment" MIR lint.
I had to whitelist some rvalues, as they are actually allowed to alias, like `a = a + 1`.
Enable DestinationPropagation by default
This PR proposes to perform destination propagation on MIR. Most of the pass was fully rewritten by `@JakobDegen` in rust-lang/rust#96451.
This pass is quite heavy, as it needs to perform and save the results of a full liveness dataflow analysis. This accounts for ~50% of the pass' runtime.
Perf sees a few decent savings in later llvm passes, but also sizeable régressions when there are no savings to balance this pass' runtime.
tests/run-make: Update list of statically linked musl targets
All of the tier 3 targets in the list now link dynamically by default (except `mips64el-unknown-linux-muslabi64`, I apparently overlooked that one in my PR that changed this).
Adjust the list of targets expected to link statically accordingly.
See also https://github.com/rust-lang/rust/pull/144410, which changed these targets.
Target by target:
- `mips64-unknown-linux-musl`: this target does not exist AFAICT
- `mips64-unknown-linux-muslabi64`: updated in the linked PR
- `powerpc-unknown-linux-musl`: updated in the linked PR
- `powerpc-unknown-linux-muslspe`: updated in the linked PR
- `powerpc64-unknown-linux-musl`: updated in the linked PR
- `riscv32gc-unknown-linux-musl`: updated in the linked PR
- `s390x-unknown-linux-musl`: updated in the linked PR
- `thumbv7neon-unknown-linux-musleabihf`: updated in the linked PR
Detect attempt to use var-args in closure
```
error: unexpected `...`
--> $DIR/no-closure.rs:11:14
|
LL | let f = |...| {};
| ^^^ not a valid pattern
|
= note: only `extern "C"` and `extern "C-unwind"` functions may have a C variable argument list
error: unexpected `...`
--> $DIR/no-closure.rs:16:17
|
LL | let f = |_: ...| {};
| ^^^
|
= note: only `extern "C"` and `extern "C-unwind"` functions may have a C variable argument list
```
Fixrust-lang/rust#146489, when trying to use c-style var-args in a closure. We emit a more targeted message. We also silence inference errors when the pattern is `PatKind::Err`.
Remove Rvalue::Len again.
Now that we have `RawPtrKind::FakeForPtrMetadata`, we can reimplement `Rvalue::Len` using `PtrMetadata(&raw const (fake) place)`.
r? ``@scottmcm``
StateTransform: Do not renumber resume local.
MIR parameters are not explicitly assigned-to when entering the MIR body. If we want to save their values inside the coroutine state, we need to do so explicitly.
This was done by renaming the `_2` local, and introducing an explicit assignment pre-transform. This particular trick confuses me.
This version makes explicit that we are assigning parameters to saved locals.
r? ``@dingxiangfei2009``
don't apply temporary lifetime extension rules to non-extended `super let`
Reference PR: rust-lang/reference#1980
This changes the semantics for `super let` (and macros implemented in terms of it, such as `pin!`, `format_args!`, `write!`, and `println!`) as suggested by ````@theemathas```` in https://github.com/rust-lang/rust/issues/145784#issuecomment-3218658335, making `super let` initializers only count as [extending expressions](https://doc.rust-lang.org/nightly/reference/destructors.html#extending-based-on-expressions) when the `super let` itself is within an extending block. Since `super let` initializers aren't temporary drop scopes, their temporaries outside of inner temporary scopes are effectively always extended, even when not in extending positions; this only affects two cases as far as I can tell:
- Block tail expressions in Rust 2024. This PR makes `f(pin!({ &temp() }))` drop `temp()` at the end of the block in Rust 2024, whereas previously it would live until after the call to `f` because syntactically the `temp()` was in an extending position as a result of `super let` in `pin!`'s expansion.
- `super let` nested within a non-extended `super let` is no longer extended. i.e. a normal `let` is required to treat `super let`s as extending (in which case nested `super let`s will also be extending).
Closesrust-lang/rust#145784
This is a breaking change. Both static and dynamic semantics are affected. The most likely breakage is for programs to stop compiling, but it's technically possible for drop order to silently change as well (as in rust-lang/rust#145784). Since this affects stable macros, it probably would need a crater run.
Nominating for discussion alongside rust-lang/rust#145784: ````@rustbot```` label +I-lang-nominated +I-libs-api-nominated
Tracking issue for `super let`: rust-lang/rust#139076
initial implementation of the darwin_objc unstable feature
Tracking issue: https://github.com/rust-lang/rust/issues/145496
This feature makes it possible to reference Objective-C classes and selectors using the same ABI used by native Objective-C on Apple/Darwin platforms. Without it, Rust code interacting with Objective-C must resort to loading classes and selectors using costly string-based lookups at runtime. With it, these references can be loaded efficiently at dynamic load time.
r? ```@tmandry```
try-job: `*apple*`
try-job: `x86_64-gnu-nopt`
remove FIXME block from `has_significant_drop`, it never encounters inference variables
The `FIXME` block in `Ty::has_significant_drop` is outdated as related queries can now handle type inference.
321a89bec5/compiler/rustc_middle/src/ty/util.rs (L1378-L1389)Closesrust-lang/rust#86868 (other places mentioned in the issue have been resolved, or moved to other issues)
r? types
Update the minimum external LLVM to 20
With this change, we'll have stable support for LLVM 20 and 21.
For reference, the previous increase to LLVM 19 was rust-lang/rust#139275.
cc ```@rust-lang/wg-llvm```
r? nikic
Rollup of 5 pull requests
Successful merges:
- rust-lang/rust#146442 (Display ?Sized, const, and lifetime parameters in trait item suggestions across a crate boundary)
- rust-lang/rust#146474 (Improve `core::ascii` coverage)
- rust-lang/rust#146605 (Bump rustfix 0.8.1 -> 0.8.7)
- rust-lang/rust#146611 (bootstrap: emit hint if a config key is used in the wrong section)
- rust-lang/rust#146618 (Do not run ui test if options specific to LLVM are used when another codegen backend is used)
r? `@ghost`
`@rustbot` modify labels: rollup
Keep space if arg does not follow punctuation when lint unused parens
Fixesrust-lang/rust#138234
If the arg follows punctuation, still pass `left_pos` with `None` and no space will be added, else then pass `left_pos` with `Some(arg.span.lo())`, so that we can add the space as expected.
And `emit_unused_delims` can make sure no more space will be added if the expr follows space.
---
Edited:
Directly use the `value_span` to check whether the expr removed parens will follow identifier or be followed by identifier.
Bump rustfix 0.8.1 -> 0.8.7
This commit can be replicated by running `cargo update -p rustfix --precise 0.8.7 && x test ui --bless`.
---
The reasons this affects UI tests is as follows:
- The UI test suite runs rustc with `-Z deduplicate-diagnostics=no --error-format=json`, which means that rustc emits multiple errors containing identical suggestions. That caused the weird-looking code that had multiple `X: Copy` suggestions.
- Those suggestions are interpreted not by rustc itself, but by the `rustfix` library, maintained by cargo but published as a separate crates.io library and used by compiletest.
- Sometime between rustfix 0.8.1 and 0.8.7 (probably in rust-lang/cargo#14747, but it's hard to tell because rustfix's versioning doesn't match cargo's), rustfix got smarter and stopped applying duplicate suggestions.
Update rustfix to match cargo's behavior. Ideally, we would always share a version of rustfix between cargo and rustc (perhaps with a path dependency?), to make sure we are testing the behavior we ship. But for now, just manually update it to match.
Note that the latest version of rustfix published to crates.io is 0.9.1, not 0.8.7. But 0.9.1 is not the version used in cargo, which is 0.9.3. Rather than trying to match versions exactly, I just updated rustfix to the latest in the 0.8 branch.
Display ?Sized, const, and lifetime parameters in trait item suggestions across a crate boundary
context: rust-lang/rust#145929
This fixes the MetaSized issue and adds const generics and early bound lifetimes. Late bound lifetimes are harder because they aren't returned by `generics_of`. I'm going to look into it, but there's no guarantee I'll be successful.
Fixes https://github.com/rust-lang/rust/issues/146404.
r? `@BoxyUwu`
When writing something like the expression `|_: ...| {}`, we now detect the `...` during parsing explicitly instead of relying on the detection in `parse_ty_common` so that we don't talk about "nested `...` are not supported".
```
error: unexpected `...`
--> $DIR/no-closure.rs:6:35
|
LL | const F: extern "C" fn(...) = |_: ...| {};
| ^^^
|
= note: only `extern "C"` and `extern "C-unwind"` functions may have a C variable argument list
```
Migrate `UnsizedConstParamTy` to unstable impl of `ConstParamTy_`
Now that we have ``#[unstable_feature_bound]``, we can remove ``UnsizedConstParamTy`` that was meant to be an unstable impl of stable type and ``ConstParamTy_`` trait.
r? `@BoxyUwU`
rustc_codegen_llvm: Adjust RISC-V inline assembly's clobber list
Despite that the `fflags` register (representing floating point exception flags) is stated as a flag register [in the reference](https://doc.rust-lang.org/reference/inline-assembly.html#r-asm.rules.preserved-registers), it's not
in the default clobber list of the RISC-V inline assembly and it would be better to fix it.
tests/codegen-llvm: Make rust-abi-arch-specific-adjustment portable
This test currently only runs on RISC-V and loongarch hosts, but assumes that the host target is the -gnu target. By using minicore, we can run this test on all host targets, regardless of architecture, as long as the LLVM components are built.
This also fixes this test on musl hosts of these architectures (though I've only tested on loongarch64-unknown-linux-musl).
